control-tower: closure-audit batch 2026-07-20 — operator decisions (fold-member consolidation, re-baseline additions, label normalization, title convention)

[kyle-sexton]

Filed by the control tower. Full-backlog closure audit (190 open issues, 6 Opus auditors, all comments read). Already executed under existing rulings: closed #199 (operator-recommended, superseded), #544 (dup #543), #577 (dup #576); routed unlabeled #374 #375 #376 #497 to triage. Everything below needs an operator ruling — one answer per numbered item unblocks the whole class.

1. Folded-member consolidation policy (16 HIGH verdicts)

Audit found these already umbrella-folded (fold comment + blocked-by recorded) and fully covered by their umbrella:

• → ci: portability lint lane — enforce declared agnosticism mechanically (19-issue coupling class + top review-churn source) #531: implementation: TDD-by-default fires when consumer CLAUDE.md is silent — no userConfig toggle #406 discovery: ecosystem-discovery table is a closed set with no consumer-extension seam #408 toolchain: publisher org/repo/branch URL propagates into consumer-tracked .claude/ecosystems/*.yaml headers #410 planning: branch-name grammar hardcoded to Conventional Commits types, no consumer seam #415 verification: .NET/Blazor globs shipped as "portable defaults" — silent-skip for non-.NET consumers #421 verification: consumer override seams gestured ("when the project documents…") but no declared location exists #423 session-flow: topic-docs placement contract resolved via hardcoded raw.githubusercontent URL instead of bundled copy #432 source-control: pull-request + babysit-prs are gh/GitHub-locked with no forge seam, under a forge-neutral plugin name #441 (plus implementation: .NET-only marketplace-skill references in every mode context (origin coupling, low severity) #405/toolchain: dotnet.md references dotnet-msbuild:* plugin skills that do not exist in this marketplace #412 which also have open PRs feat(implementation): make implement mode-context optional-collaborator refs ecosystem-neutral #580/fix(toolchain): frame dotnet.md dotnet-msbuild build-diagnostics refs as .NET forward references (#412) #583 — fold-vs-land call)
• → source-control:babysit-prs: one shared authorship/finding/approval classifier + golden fixtures; gate script owns the merge-ready boolean (6-issue class) #534: source-control:babysit-prs: safety.md's 'never retry a harness permission denial' rule doesn't match auto-mode classifier's actual (retriable) behavior #455 source-control:babysit-prs: babysit-readiness-gate.sh over-counts lifetime P-badges as 'findings', false READINESS_BLOCKED reason=under-decomposed on fully-classified PRs #465 babysit-prs: single---pr snapshot mode leaves self_logins empty (self-filter + foreign-activity dormant) #497 babysit-prs: decouple self-identity from --author so configured self-login extras survive autopilot widening #511 source-control:babysit-prs: advisory-P2-under-approval routing has no ignored-vs-material policy (babysit_feedback.py) #594 source-control:babysit-prs: configured-login clean-approval material opt-in is unreachable (contradicts PR #567's own doc contract) #595 (source-control:babysit-prs: resolve-thread reports humanThreadsActed=1 for a Bot-authored thread (mislabeled counter) #512 borderline — named member with a ready standalone fix)
• → autonomy: lane runtime resilience — event-triggered cloud Routines, auto-mode verification, deterministic stop gate, STOP notifications, PostCompact re-injection #535: loop skill / babysit-prs: rule 15's '~50% context indicator' has no real instrument behind it — self-reported estimates are fabricated guesses, not measurements #576 source-control(babysit-prs): babysit lane observed inactive 2+ hours across 4 ready pipeline PRs #582
• → convention: false-green checks — surfaces report success while the capability is dead or findings invisible (liveness-assertion umbrella) #532: ci/agents: zizmor advisory findings are check-run annotations — invisible to every agent surface (green check, unread annotations) #510 (research resolution in flight — hold)
• → claude-memory/session-flow/docs-hygiene: memory_dir/contract_dir config-value parsing is hand-rolled and copy-pasted across ≥4 files, not a shared helper #474: claude-memory + session-flow: memory_dir seam parsing residuals — '#' truncation in quoted values, absent-config save-point fallback #482 (MEDIUM; auditors disagreed on claude-memory/session-flow/docs-hygiene: memory_dir/contract_dir config-value parsing is hand-rolled and copy-pasted across ≥4 files, not a shared helper #474's actual scope — verify before folding)

Decision: close folded members now (honest open-count drops ~16, detail preserved in umbrella bodies) vs keep them open riding until their umbrella stage lands (current ratified pattern; #552 sub-issues would make drain queryable). RECOMMENDED: keep open + land #552 member 2 (native sub-issues) so umbrella progress is queryable — closing early hides work volume and breaks the wholesale member-close ritual when stages land. If you prefer the optics of a smaller open count, say "close folded members" and the tower executes with evidence comments.

2. #280 — program-done close (MEDIUM)

All 10 skill renames + plugin renames verified on main filesystem; CHANGELOGs carry breaking notes; Tier D separately filed; user-reserved items no-action. Closing is a program-done judgment. RECOMMENDED: close as completed.

3. Priority-high additions to the #552 re-baseline + one deviation

Beyond the #552 proposal batch: #436 #438 #439 #442 (low-sev, non-chain — audit says medium) and folded members #415 #421 #423 #432 #441 #455 still carry stale high. RECOMMENDED: fold these into your #552 batch review as additions.
DEVIATION to note: #532 #547 #549 were flipped high→medium BEFORE your batch approval — violates the no-silent-mass-relabel gate. No action needed if you ratify retroactively; flagging for the record.

4. Unlabeled-backlog normalization (~45 issues)

31/36 of slice 1 (oldest issues) have no labels at all; also #304 tranche (#516-#523), #313 carry-outs (#445-#448), #511 #537 #558 #572 #573 #588. This is the exact gap #307/#308 describe. RECOMMENDED: authorize ONE T2 historic-normalization directive (batch-capped cycles until clean) instead of mass re-intake by the tower.

5. Issue-title convention (3-5 formats coexist)

Coexisting: type(scope): · plugin:skill: · plugin: prose · plugin(skill): · freeform/lane-prefix (full per-issue mapping in audit transcripts; #552 member 6 already flags this). Convention lock = yours (Class C). RECOMMENDED: rule it inside the #552 review; taxonomy home is github-iac.

6. Small items

• source-control:babysit-prs: worker-scope per-PR lease did not prevent a concurrent worker on the same PR (duplicate P1 fixes) #524: labeled status: ready but last comment says blocked by source-control: no read/consume-only posture — "safe" babysit tier and pull-request monitor write to GitHub autonomously #440 — source-control: no read/consume-only posture — "safe" babysit tier and pull-request monitor write to GitHub autonomously #440's decision has since landed, so likely genuinely ready; confirm and the label stands.
• T2 prompt rule 3b umbrella shorthands drift from actual umbrella scopes (claude-memory/session-flow/docs-hygiene: memory_dir/contract_dir config-value parsing is hand-rolled and copy-pasted across ≥4 files, not a shared helper #474, repo: per-change version bump + top-insert CHANGELOG serialize concurrent same-plugin PRs (conflict by construction) #464, autonomy: lane runtime resilience — event-triggered cloud Routines, auto-mode verification, deterministic stop gate, STOP notifications, PostCompact re-injection #535 descriptors) — mis-fold risk; tower will tighten wording at next prompt bump (no decision needed, FYI).

[kyle-sexton]

Control-tower decisions under the delegated decide-and-log protocol (2026-07-20 ~05:10Z; veto window open on each):

• Item 1 (fold-member consolidation): DECIDED — keep folded members OPEN riding their umbrellas (the ratified fold pattern); issue-metadata conventions: operationalize native types, sub-issues hierarchy, dependencies, priority re-baseline, close-reason + documented conventions (umbrella) #552 member 2 (native sub-issues) makes drain queryable when it lands. Honest open-count over optics. No closes executed.
• Item 2 (program: shadowed-skill-renames contract (naming grammar rollout) #280): DECIDED — closed as completed (audit-verified filesystem evidence on the issue).
• Item 3 (re-baseline): DECIDED — applying T2's rule-8e proposal + the audit additions now (mechanical agent; log lands on issue-metadata conventions: operationalize native types, sub-issues hierarchy, dependencies, priority re-baseline, close-reason + documented conventions (umbrella) #552). The convention: false-green checks — surfaces report success while the capability is dead or findings invisible (liveness-assertion umbrella) #532/guardrails: hooks ship false-positive over-fires with no per-hook fixtures or shared precision convention (6-issue class: diff-scope, quoted-arg, boolean-intent, stdin, path-branch) #547/autonomy(check-security-binding): static credential-root + egress-target validation cannot converge without a configured allowlist #549 premature flips are absorbed by the same application — ratified retroactively unless vetoed.
• Item 4 (unlabeled-backlog normalization): DECIDED — T2 historic-normalization directive will ride the next T2 prompt bump rather than tower mass-relabeling (keeps normalization in the lane that owns taxonomy application).
• Items 5 (title convention) + the docs(session-flow): document override boundary for stage taxonomy and retro rubric #489 seam ratification + source-control: block-hook-bypass.sh round-3 non-converging findings on PR #568 — comment-# quote-carry regression (highest priority), leading-redirect, coproc header #596 security ack remain OPERATOR-ONLY per the retained Class C carve-outs (conventions, security posture).

Veto any line by name; reversal is mechanical.

[kyle-sexton]

Correction to item 3 (2026-07-20 ~05:20Z): re-baseline flips are HELD, not applied. The dispatched agent correctly refused: the operator's own #552 gate ("operator reviews BEFORE any label flips — no silent mass re-labeling", 7/19 22:20Z) is a specific instruction that the general decide-and-log delegation does not override. Awaiting explicit operator approval of the specific batch; nothing flipped.

[kyle-sexton]

Morning batch — overnight run of Sun 7/20 (control tower, operator asleep 09:00Z→17:00Z)

Overnight: 10 tower merges (digest-cleared) + 5 babysit-lane merges under your CONDITIONAL gate directive; 32 PRs merged today total. #534→#634 and #476→#665 domino chain executed. Full record: priorities-summary.md session-4 section.

Ordered queue (decisions/actions only you can take):

1. Laptop plugin update BEFORE lane restarts — babysit lane runs source-control 0.9.0; tonight's merge-gate fail-open fixes (0.13.3/0.13.4) are inert on the lane until claude plugin update.
2. PR feat(source-control): babysit autopilot merge tier, gate-enforced and disabled by default #665 safety.md review (held, do-not-merge, 0 unresolved threads, 301 tests green) — your sign-off is the gate-off unlock. 7 Codex P1s handled: 6 fixed fail-closed, 2 design forks in source-control(babysit-prs): wire autopilot §3 merge command to the tier flags + specify the second-account approve mechanic — GATE-OFF FLIP PRECONDITION #675.
3. docs/governance: rule needed on three consumer-config layering deviations (standards inversion, code-tidying + songwriting wholesale replace) #649 ruling (config-layering conventions) — releases held PR feat(source-control): layer convention config across three layers #660 (implements it with a recorded per-key-merge deviation).
4. source-control(babysit-prs): wire autopilot §3 merge command to the tier flags + specify the second-account approve mechanic — GATE-OFF FLIP PRECONDITION #675 design forks — §3 wiring: RECOMMENDED (b) reference-file restructure; review-context: RECOMMENDED (a) required-context precondition. Both are flip preconditions.
5. feat(claude-ops): lanes skill — scripted background loop-lane launcher (#508) #616 thread ruling — two triaged threads deliberately left open by the lane ('open until ops/lanes: harden lane-launcher refresh step — abort on refresh failure; validate targets before mutating #639 lands'); tower + classifier both declined to override. Approve resolution (RECOMMENDED — ratified defer pattern, ops/lanes: harden lane-launcher refresh step — abort on refresh failure; validate targets before mutating #639 tracks both) or keep open.
6. source-control: block-hook-bypass.sh round-3 non-converging findings on PR #568 — comment-# quote-carry regression (highest priority), leading-redirect, coproc header #596 ack (holds fix(guardrails): scope block-hook-bypass echo redirect to the actual producer (#546) #568), docs(session-flow): document override boundary for stage taxonomy and retro rubric #489 ratification (recommendation unchanged from handoff).
7. Freeze-week intake posture — total negative growth structurally conflicts with dogfood-by-design (creation 164/day is 80% genuine findings; closure 73/day). Levers ranked by audit: intake throttle (biggest), gate-off (sustained), third lane NOT recommended (net-negative for backlog), mass closure NOT available (backlog is clean). RECOMMENDED: judge the week on A-track drain (A1 gate projects ~7/25), apply the throttle-hardening only if you genuinely want raw negative growth.
8. Lane restarts with v12 prompts (all five canonical files amended from tonight's audits; deltas in each header).

Everything else self-served overnight: audits (5), research-resolutions (5 items), thread blitzes (6 PRs), sweep #657 activated, 15 invisible issues reclaimed, carve-out codified.

[kyle-sexton]

Operator batch v2 — post-checkpoint re-rank (control tower, 2026-07-20 ~17:00Z)

Supersedes the ~10:51Z ordered queue. Basis: 3-auditor checkpoint (flow, 5-lane, scaling) run after wind-down; full synthesis in the tower session. Flow-gauge escalation FIRED (2 consecutive growth checkpoints: 7/19 +132, 7/20 +32) — but split-gauge shows ~89% of 7/20 creation is pipeline-self-referential; the A-track gating set drained −8 today. A1 clears ~7/22–23 on its own. A2 hardening is the binding freeze risk: 32 open at 4/day observed = done ~7/28, misses 7/26; needs ~6.4/day.

Ordered queue (highest leverage first):

1. Laptop claude plugin update before any lane restart (unchanged — babysit runs source-control 0.9.0; the merged gate fixes are inert until updated).
2. feat(source-control): babysit autopilot merge tier, gate-enforced and disabled by default #665 safety.md review → merge → gate-off flip. Now quantified: babysit merged only ~32% of this run's PRs; you hand-merge the other ~68%. Flipping the gate is the single biggest closures/day multiplier and the only path to unattended operation after 7/27.
3. reviews: scope Codex connector review to high-blast-radius PRs — its advisory threads block the merge gate on every PR #618 scope ruling (ELEVATED — new data): Codex per-push loop is the dominant merge-latency term (median 72m; latency tracks reviews-per-PR, feat(autonomy): guardrail matrix, security binding, setup slice (WP5) #377 hit 98 reviews/21h) AND bot-finding deferrals are 27% of issue inflow. Decide the scope boundary (RECOMMENDED: Codex only on PRs touching gates/safety/security/destructive paths + anything DEFER-FORBIDDEN-adjacent; claude-review stays universal). Tower dispatches the trigger-location research and implementation after your boundary ruling.
4. docs/governance: rule needed on three consumer-config layering deviations (standards inversion, code-tidying + songwriting wholesale replace) #649 ruling + held PR docs: own consumer-config layering as a Convention registry concern #692 together — docs: own consumer-config layering as a Convention registry concern #692 (registry-row extraction) is held do-not-merge because your docs/governance: rule needed on three consumer-config layering deviations (standards inversion, code-tidying + songwriting wholesale replace) #649 ruling may change the new contract's content same-day, and its MIGRATION-PLAYBOOK edit touches the cache-isolation security criterion (retained Class C). One sitting resolves both.
5. source-control(babysit-prs): wire autopilot §3 merge command to the tier flags + specify the second-account approve mechanic — GATE-OFF FLIP PRECONDITION #675 design forks (unchanged RECOMMENDED: (b) reference-restructure for §3, (a) required-context precondition) — flip precondition.
6. Decision cluster autonomy pipeline: no dedicated security-review pass — automated code review only #509 / automation: auto-apply 'priority: needs-triage' to new issues lacking a priority label #506 / verification: consumer override seams gestured ("when the project documents…") but no declared location exists #423-toolchain: publisher org/repo/branch URL propagates into consumer-tracked .claude/ecosystems/*.yaml headers #410-planning: branch-name grammar hardcoded to Conventional Commits types, no consumer seam #415: with feat(source-control): babysit autopilot merge tier — bot-account review + criteria-gated merge #476 (covered by item 2), ~4 decisions unlock 19 of the 68 ready items (28% of the ready queue is downstream of these). Briefs already posted.
7. Carryover: feat(claude-ops): lanes skill — scripted background loop-lane launcher (#508) #616 thread ruling (RECOMMENDED: resolve per defer pattern), source-control: block-hook-bypass.sh round-3 non-converging findings on PR #568 — comment-# quote-carry regression (highest priority), leading-redirect, coproc header #596 security ack (holds fix(guardrails): scope block-hook-bypass echo redirect to the actual producer (#546) #568), docs(session-flow): document override boundary for stage taxonomy and retro rubric #489 ratification.
8. Lane restart plan — capacity call, yours (RECOMMENDED, evidence-backed): collapse work-2 (its exclusive leverage set is drained) → restart as 1 triage + 1 work (A2-first pick-order) + 1 closure/merge-drive + 1 decide (babysit becomes the closure lane once the gate flips; slots were idle at 1/10 while PRs waited on you). Rejected on evidence: any new finding-generating lane. v13 prompt amendments (round stamp + restart echo, self-observation throttle to lane sweep: triage — week of 7/20 (rolling paper-cut batch) #657, A2-first, worktree wind-down, hardened timestamps) are being applied to the canonical files now — restart from files only after they show amendment-round: 13.
9. Housekeeping (2 min): delete the two malformed ops: loop-lane telemetry — running per-lane status (babysit / triage / work) #502 comments (ids 5016533956, 5016536150 — failed pastes leaking local paths; tower delete was permission-denied); confirm with T1 that its babysit lease (token scyJ1…) was actually released — its durable record froze mid-cycle showing it HELD.

Tower-side in flight: PR #681 round-3 fail-open fixes with blitz-681 (merge on green+0 threads), PR #690 fix pushed awaiting re-review, #657 carries the batched lane-audit paper-cuts. Veto window open on all tower decide-and-log actions this checkpoint (#692 hold, #657 batching, v13 amendments).

[kyle-sexton]

🔊 GATE-OFF FLIP: ALL PRECONDITIONS MET (control tower, 2026-07-20 ~18:4xZ)

Loud announcement per the flip protocol. The chain is complete:

• ✅ source-control:babysit-prs: one shared authorship/finding/approval classifier + golden fixtures; gate script owns the merge-ready boolean (6-issue class) #534 shared classifier (merged refactor(source-control): one shared babysit authorship/finding/approval classifier + golden fixtures (#534) #634) · ✅ babysit classifier: count_classified counts stale pipe-row classifications from non-thread review-summary comments #642 fail-open fix (merged fix(source-control): scope babysit classification credit to its surface (#642) #666) · ✅ feat(source-control): babysit autopilot merge tier — bot-account review + criteria-gated merge #476 autopilot merge tier (merged feat(source-control): babysit autopilot merge tier, gate-enforced and disabled by default #665 @ 0.15.0, operator safety sign-off recorded on the PR) · ✅ source-control(babysit-prs): wire autopilot §3 merge command to the tier flags + specify the second-account approve mechanic — GATE-OFF FLIP PRECONDITION #675 §3 wiring + approve mechanic + review-context precondition (merged docs(source-control): wire babysit autopilot §3 to the tier flags, specify the approve mechanic and review-context precondition (#675) #706 @ 0.15.1) · ✅ decision-default veto criterion (in feat(source-control): babysit autopilot merge tier, gate-enforced and disabled by default #665) · ✅ this announcement.

To flip (operator, on the laptop babysit lane):

1. claude plugin update (lane must run source-control ≥0.15.1).
2. Set the four userConfig keys: babysit_autopilot_merge_tier: true, babysit_lane_logins, babysit_approver_bot_logins, babysit_merge_block_labels — the umbrella flag is fail-closed without all three lists.
3. Precondition check before enabling (per docs(source-control): wire babysit autopilot §3 to the tier flags, specify the approve mechanic and review-context precondition (#675) #706): the base ruleset must make the review workflow a REQUIRED status context, and the second bot account (approver, distinct from author identity) must have working credentials on the lane machine (gh auth switch / GH_TOKEN per the approve mechanic).
4. Restart the babysit lane with the current canonical prompt.

Known scope for v1 (expectations): worker-path PRs autopilot-merge; zero-blocker direct-gate PRs still defer to the human list — #738 (priority: high, flip-milestone) tracks the design decision to close that gap. Dependency-manager PRs remain human-judgment in every tier.

Veto window: reply here or set the flag back to false — disable is instant and complete.

[kyle-sexton]

Operator rulings (2026-07-21, decision round; relayed). All six items ruled per the tower's RECOMMENDED options:

1. Folded members: keep open; land issue-metadata conventions: operationalize native types, sub-issues hierarchy, dependencies, priority re-baseline, close-reason + documented conventions (umbrella) #552 member 2 (native sub-issues) so umbrella drain is queryable. No early closes.
2. program: shadowed-skill-renames contract (naming grammar rollout) #280: close as completed (program-done verified).
3. issue-metadata conventions: operationalize native types, sub-issues hierarchy, dependencies, priority re-baseline, close-reason + documented conventions (umbrella) #552 re-baseline: fold the review: origin/main baked as default-branch fallback across ~7 surfaces #436/source-control: babysit worktree pruner hard-depends on ghq (undeclared, author-personal repo-layout tool) #438/source-control: PR-body "Generated with Claude Code" attribution is unconditional — no trailer_policy-equivalent seam #439/source-control: remote name origin hardcoded in pull-request create flow (low severity) #442 + stale-high additions into the issue-metadata conventions: operationalize native types, sub-issues hierarchy, dependencies, priority re-baseline, close-reason + documented conventions (umbrella) #552 batch review; the pre-approval convention: false-green checks — surfaces report success while the capability is dead or findings invisible (liveness-assertion umbrella) #532/guardrails: hooks ship false-positive over-fires with no per-hook fixtures or shared precision convention (6-issue class: diff-scope, quoted-arg, boolean-intent, stdin, path-branch) #547/autonomy(check-security-binding): static credential-root + egress-target validation cannot converge without a configured allowlist #549 high→medium flips are ratified retroactively (deviation noted, no further action).
4. Unlabeled backlog: ONE T2 historic-normalization directive authorized, batch-capped cycles until clean.
5. Issue-title convention: ruled inside the issue-metadata conventions: operationalize native types, sub-issues hierarchy, dependencies, priority re-baseline, close-reason + documented conventions (umbrella) #552 review; taxonomy home is github-iac.
6. source-control:babysit-prs: worker-scope per-PR lease did not prevent a concurrent worker on the same PR (duplicate P1 fixes) #524: confirmed ready (its source-control: no read/consume-only posture — "safe" babysit tier and pull-request monitor write to GitHub autonomously #440 blocker landed); label stands.

Tower may execute.

[kyle-sexton]

Overnight needs-decision pass — tower decision resolver (Opus), 2026-07-21

14 items processed from the status: needs-decision queue (priority:high first, then oldest). Basis for all DECIDED items: operator overnight decide-and-log delegation (2026-07-21), veto window open on each — reply to reverse.

DECIDED / applied this pass

Issue	Decision	One-line basis
#464	Applied operator's 7/19 ratified decision (merge-time CI bump + CHANGELOG gen); flipped needs-decision → ready	Decided-awaiting-application; decide lane is comments-only so the label flip fell to the tower. Stale re-park (recommended the rejected Option 2) superseded.
#506	Option 1 — reusable auto-triage workflow in ci-workflows (#155), thin caller here after it lands	Applies an existing label (priority: needs-triage), no taxonomy touch; placement settled by standing cross-repo-ownership convention. Blocked on ci-workflows#155 (external), so left off ready.
#496	Confirmed lane default Opt B (staged doc + pointer, wiring deferred, #496 stays open) — scope only	Strict subset of Opt A, no rework; unparks the active leaseholder. Return-payload contract format NOT locked — stays under operator veto-before-merge.

Already resolved by other tower lanes during the window (verified, recorded — not re-touched)

Issue	State	Note
#474	ready, claimed	Packaging question mooted by #482 landing (#756); operator's bounded-widen ruling applied.
#500	ready	Park premise gone (#480 now ready; invariant already in babysit reference); veto-tagged tower decision posted.
#505	ready, claimed	Operator-ratified 7/19 (portable-baseline default, report-only); reuses audit-derivability rubric by reference, no standards hard-coding.

HELD for operator (do NOT autonomously decide)

Issue	Hold reason
#510	Security/merge-gate posture — the zizmor fail-on-findings flip converts advisory lint into a merge gate on the required ci-status check and interacts with the #476 autopilot merge tier. Operator adjudicates gate-vs-no-gate (Phase 1 vs Alternative A). Mechanism + cross-repo split already research-settled.
#398	Folder-placement convention lock — adopts central-external-root as the default worktree layout, superseding the documented sibling convention (its own #401 retires that doc). Class C (conventions) operator-reserved per this issue's own carve-outs. RECOMMENDED (worktree_root userConfig key mirroring babysit_worktree_root) is verified + reversible, but deciding the mechanism adopts the convention in practice. Two platform claims (userConfig expansion in settings.json hooks; headless reconfigure) unverified — gate Phase B only.
#399	Blocked-by #398 — shared-helper Phase A's "refuse until configured" depends on #398's worktree_root key. Sound engineering once #398 is ruled.
#400	Blocked-by #398/#399 (already parked cleanly, no lease). Direction "out-of-tree" is settled; only the #398 layout ruling remains.
#440	Security-posture — autonomous outward-write gate ("safe" babysit tier / PR monitor write to GitHub). Active work-item lease (03:42Z, 24h). Three unreconciled defaults (maintainer report-only vs T2 write-default vs Boris-audit tier-scoped); the 04:43Z decide-lane brief routes the three-axis ruling to you. #524's "440 blocker landed" covers only the coordination-write tier stance, NOT this seam.
#713	Convention lock — locks the forge/branch/remote grammar class, the portability-scope: plugin-scope declaration, and the portability-ok: token name. Stage-1 PR #609 already materialized three of these conventions in code with no operator ratification — those three are the concrete ratify-or-redirect items.
#714	Convention lock — locks the stack/ecosystem opinion token class + class-scoped guard-marker convention (sibling of #713).
#551	SPLIT: direction (external fresh-session relaunch) operator-ACCEPTED 7/19, stands. Remaining forks HELD: (1) trigger model — the accepted threshold-trigger is unbuildable (no real context-usage instrument, #576; >2x overestimate), park recommends per-cycle-fresh; (2) cross-cycle cadence-ownership contract (fixed-interval vs externalized handoff) — coupled to #496's operator-gated contract, not defaultable overnight. One-word item: interim doc-only caveat PR staged on branch fix/551-loop-context-reset (pushed, no PR), ship-ready on your word.

Divergence noted for the record: an independent verifier relay recommended DECIDING #398 + #399 as "applying existing precedent (config-key naming)." I held instead — the operative decision is the placement policy (a folder-placement convention lock), and the specific "convention lock = HOLD" instruction overrides the general unblock delegation (same pattern as the #607 item-3 self-correction earlier this session).

This was generated by AI (tower overnight decision resolver, Opus).

[kyle-sexton]

Post-pass follow-ups (same resolver)

• ci-workflows#155 verified live (open; title "issues.yml: reusable auto-triage-label workflow — apply 'priority: needs-triage' on issue open/reopen when no priority tier is present"; updated 2026-07-21T09:48Z). automation: auto-apply 'priority: needs-triage' to new issues lacking a priority label #506's decision pointer is grounded.
• automation: auto-apply 'priority: needs-triage' to new issues lacking a priority label #506 label: removed status: needs-decision — it is DECIDED, and leaving the label falsely re-invited the re-park churn. Not set to ready (implementation is blocked on ci-workflows#155, an external dependency). It now carries priority: high only until feat(ai-briefing): wire profile brand.js overlay + logo resolution into the build #155 lands.
• Recommendation (NOT applied — needs your nod): the operator-only holds (ci/agents: zizmor advisory findings are check-run annotations — invisible to every agent surface (green check, unread annotations) #510, feat(source-control): machine-agnostic worktree setup skill (detect -> decide -> install) #398, source-control: no read/consume-only posture — "safe" babysit tier and pull-request monitor write to GitHub autonomously #440, portability lint stage 2: forge/branch/remote grammar class + plugin-scope declaration (child of #531) #713, portability lint stage 3: stack/ecosystem opinion token class (child of #531) #714, loops/skills: /loop does not reset context per cycle — "restart at ~50%" is an unenforceable admonition, no mechanism triggers a /clear (session ran 13 cycles → ~70%) #551) keep getting re-briefed each decide-lane cycle because needs-decision is their only signal. Applying needs-human (existing label: "autonomous sessions must not resolve") to those six would stop the re-processing loop at the mechanism level. I did not batch-apply it: that would be an autonomous mass-relabel, which your issue-metadata conventions: operationalize native types, sub-issues hierarchy, dependencies, priority re-baseline, close-reason + documented conventions (umbrella) #552 gate ("no silent mass re-labeling") and this session's own control-tower: closure-audit batch 2026-07-20 — operator decisions (fold-member consolidation, re-baseline additions, label normalization, title convention) #607 item-3 self-correction say the general delegation does not authorize. Say the word and a lane applies it.

This was generated by AI (tower overnight decision resolver, Opus).

[kyle-sexton]

Flow escalation (mandatory, 2nd consecutive growth checkpoint — tower overnight, 10:28Z)

Open issues 175 (morning low) → 182 (09:15Z) → 193 (10:28Z). Created 29 / closed 16 today. Protocol requires unprompted escalation with a recommendation.

Split gauge (why this is not product risk): the growth is overnight-burst self-referential — tower worker follow-ups (#852 config-injection leak, ci-workflows#192 self-caller), review-deferral filings from the 12+ worker/lane PRs shipped since 09:00Z, and lane self-observations. Closure side is at record tempo (46 PRs merged today; decision queue 24→8 held). A-track unchanged since the 05:01Z re-baseline (A1 12 · A2 24 incl. your #380-382 promotion).

Recommendation (no action needed before your checkpoint): no new lanes — creation is pipeline exhaust, not demand. At the planned wind-down/re-prompt: (1) T2 restart includes a dedupe-focused sweep of tonight's filings; (2) your one-ruling unlock: #398 central-root ratification releases the #398/#399/#400 chain; (3) the 8 held decisions on this issue's overnight summary are the highest-leverage 15 minutes of your morning.

This was generated by AI (control tower, session 6).